All products

Read a bank CSV with AI

Get a read-only profile of a bank export before asking questions about it. See which dates, signs, columns and source rows the figures depend on.

Read a bank CSV with AI before comparing it with another month’s export. A heading called “amount” does not tell you whether money leaving the account appears as a negative number or sits in a separate debit column. The read-bank-csv skill asks a deterministic tool to inspect the layout. The agent explains what the parser found and where in the source file you can check it.

Your profile records the delimiter, header row, transaction count, date range and anything the parser skipped. A preamble still counts towards the original row numbers. If both day and month could be twelve or less throughout the file, the guide asks which convention your bank uses.

What it does

  • Identify a balance column and check the direction of a withdrawal against it.
  • Keep a closing summary out of the transaction count.
  • Spot a decimal comma before treating an amount as a whole number.
  • Report an unreadable row with its location and the reason it was skipped.
  • Check dates shared by two statements before combining their figures.

Install and use

  1. Use your exported CSV and an agent with a local parser or the optional ledger-mcp tool. No online banking login is required.
  2. Ask, “Profile this export without changing it. Which columns are payments and receipts, and how did you verify that?”
  3. Follow a row citation back to the original. Settle any uncertain date or sign convention before using the figures.

FAQ

Source

Made by FolioFlux. Original skill: read-bank-csv, published on GitHub under the MIT licence.

View on GitHub